风险价值VaR的计算方法研究

发布时间：2018-01-08 17:37

本文关键词：风险价值VaR的计算方法研究　出处：《重庆大学》2013年硕士论文　论文类型：学位论文

【摘要】：近几十年来，在经济与金融逐步的全球化、一体化、现代金融理论以及金融与信息技术的创新等等诸多方面因素的影响下，随着全球金融市场迅猛发展的同时来临的还有市场所呈现出的前所未有的波动性，因此，金融风险引起了全球工商企业、金融机构等社会各方面以及学术界的密切关注。在这种背景下，风险管理能力也顺势成为经营管理的核心能力之一。作为风险管理工具的VaR(Value atRisk)是一种有效的金融风险测量手段，它最初主要用于市场风险的度量，后来逐步在国际上得到了广泛的应用和重视,并渐渐成为了金融市场风险度量的主流方法和众多投资领域中风险管理的国际标准。所以，研究如何能够更加准确的计算VaR不仅在理论上，更在实际生活中都具有十分重要的意义。风险价值的传统计算方法中往往在理论上假设其满足一些如正态分布一类的椭圆型分布，没有充分考虑到金融资产中极常出现的“厚尾”现象，所以在对分布尾部的极值拟合方面效果不佳，以致影响金融危机等极端条件下的风险预测效果。基于上述情况，论文提出了一种新的基于G-H分布的VaR计算模型，，并采用随机模拟的方法来拟合模型中的重要参数，最后通过对股票市场的数据进行实证分析，证明了该方法对VaR的拟合效果优于历史模拟法和正态分布法，并且在计算步骤与方法上面比目前使用的较多的极值理论法更为灵活和准确。本论文完成的主要工作及研究成果如下： ①充分学习了计算风险价值VaR的基本理论和相关模型，较全面地综述了国内外相关研究的现有情况，为综合研究风险价值VaR计算的方法奠定了基础。 ②系统地总结了风险价值VaR计算中的极端值拟合方法，学习了处理金融风险中对极端情况进行预测的国际前沿技术，为在对“厚尾”分布下的风险价值VaR计算方法的研究奠定了理论基础。 ③针对风险价值VaR的一般计算模型中没有充分考虑到对金融资产中极常出现的“厚尾”情况以致影响极端条件下风险预测效果的问题，通过对各项相关数学拟合模型的学习和分析，建立了一种使用G-H分布对历史数据分布进行拟合的算法。 ④针对G-H分布模型中传统参数拟合方法中难以做到使拟合所得到的四阶矩同时与目标分布一致的问题，通过对参数拟合思想、传统参数拟合算法的学习和分析，提出了一种基于随机模拟的参数拟合方法，能够做到在操作性简洁、方便的前提下保证模型的四个重要参数同时逼近历史数据的四阶矩，有效提高了模型的拟合优度，从而提高了预测VaR的准确率。从某证券公司网站提供的上证指数日收盘数中选择了一组样本数据，利用新的方法进行拟合模拟。结果显示：新的VaR计算方法不仅在置信度取值较高的情况下拟合效果优于传统的VaR计算方法，并且在操作性上较目前较为广泛应用的EVT方法更加简洁灵活。
[Abstract]:With the rapid development of global financial markets and the unprecedented volatility of the market , the risk management ability has become one of the core competencies of the management . In the traditional calculation method of risk value , it is often theoretically assumed that it satisfies some elliptic distribution such as normal distribution , which does not take fully into account the phenomenon of " thick tail " appearing in the financial assets , so as to affect the risk prediction effect under extreme conditions such as financial crisis . Based on the above , this paper proposes a new VaR calculation model based on G - H distribution , and proves that the fitting effect of this method to VaR is better than that of historical simulation method and normal distribution method . The main work and research results completed in this paper are as follows : ( 1 ) The basic theory and relevant models for calculating the VaR of risk value are fully studied , and the current situation of relevant research at home and abroad is comprehensively reviewed , which lays a foundation for the comprehensive study of VaR calculation method . In this paper , the extreme value fitting method in VaR calculation is systematically summarized , and the international forward technology for forecasting extreme cases in financial risk is studied , which lays a theoretical foundation for the study of VaR calculation method under " thick - tail " distribution . ( 3 ) In the general calculation model of VaR , the problem of " thick tail " in financial assets is not fully taken into account so as to affect the risk prediction effect under extreme conditions . Through the study and analysis of relevant mathematical fitting models , a new algorithm for fitting historical data distribution using G - H distribution is established . ( 4 ) For the traditional parameter fitting method in the G - H distribution model , it is difficult to make the fourth - order moment obtained by the fitting be consistent with the target distribution . Through the learning and analysis of the parameter fitting idea and the traditional parameter fitting algorithm , a parameter fitting method based on the stochastic simulation is put forward , which can ensure that the four important parameters of the model simultaneously approximate the fourth - order moment of the historical data under the premise of simple operation and convenience , thereby effectively improving the goodness of fit of the model so as to improve the accuracy of predicting VaR . The results show that the new VaR calculation method is superior to the traditional VaR calculation method under the condition that the confidence value is high , and is more concise and more flexible in operation than the widely used EVTs method .

【学位授予单位】：重庆大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F830.91;F224;O211.67

【参考文献】